Atmel AVR + Motorola MPX Sensors approach - Notes and Discussions to the project

[MrHijet]

[bruce]

I would go with all "Dp" series sensors so you have the ability to swap connections when you change from intake to exhaust testing, unless you plan to do this via software?

The MPX2010 series sensors will get you 40" h2o

I'm learning as we go along here also so don't take my ideas as "gospel"



Edited By bruce on 1211112190

[MrHijet]

Hi Bruce,

The MPX 2xxx series is not temperature compensated, thats why I got the hint from 86rocco1 to use the MPX 5xxx series. Additionally the 2xxx series need 10 volt versus the 5xxx series is still happy with 5 volt.

The 010 Models are made up to 10 kPa, the 050 to 50 kPa and finally the 100 model to kPa.

Eventually I am wrong, but doesn't it make sense to have a static sensor for the overall depression and a differencial sensor for the difference before and after the orifice ?

My idea was to add both to the processor.

Daniel

[bruce]

[bruce]

As a followup the Dp sensor across the orifice might only be seeing a total 8-12" h2o depending on your orifice plates.

[86rocco1]

Unless you have some means of changing the gain on the sensors, I think it's best to match, as closely as possible, the range of the sensor to the highest pressure you expect to measure otherwise, you're wasting A/D resolution. If for example the DP on your orifice is 12" or about 3Kpa and you're using a 50 Kpa sensor, you're only using about 1/16 of the full sensor range so you're effective reducing your 10 bit a/d by a factor of 16 down to 6 bits which isn't really adequate. For my bench, I was thinking of using a 10Kpa for the test pressure and a 4Kpa sensor from the MPXV5xxx series for the differential pressure.

The 5000 series sensors have on-chip amplifiers, this does make them much easier to interface with microcontroller but at a cost, you have no control over the signal gain which is why I said they're less versatile. With the 2000 series, because it requires an external signal amplifier, you have the option of setting up the signal amplifier so that the 0-3 Kpa range gives you an output signal of 0-5v even if you're using a 50Kpa sensor thus allowing you to fully exploit the a/d capability of your microcontroller.

[MrHijet]

Okay, understand.

I wasn't sure if 10kPa would be enough, as I hear of high pressure testings in the flow bench projects forum. So I will try my luck with 5010.

And to "emulate" the water gauges, I will use two differential sensors. One across the orifice and the other one to the plenum box and the other side "open".

I still have the feeling that I know far too less on the flow bench stuff ... sorry for that.

Daniel

[bruce]

[86rocco1]

[MrHijet]

Okay, lets stay with the MPX 5010 DP-NP.

Then I have 0 to 10 kPa range, which will be reflected from 0.2 to 4.7 volt.

The internal 10 bit a/d converter works fine with a resolution of 2^10 = 1024 at 5 volt, which would be a resolution of 1 bit per 4,88 mV.
The resolution of the MPX5010 is 450mv per kPa, so per kPa I would have 92 bits resolution.

Based on a 12 bit a/d, it would be 4096 bits = 1 bit per 1,22 mV.
Based on a 14 bit a/d, it would be 16384 bits = 1 bit per 0,31 mV.
Based on a 16 bit a/d, it would be 65536 bits = 1 bit per 0,08 mV.

So I could probably build it in following variations:

10 bit: resolution of 92 per kPa
12 bit: resolution of 368 per kPa
14 bit: resolution of 1451 per kPa
16 bit: resolution of 5625 per kPa

What is the typical water-gauge "resolution", when you measure a head ? How precise does it have to be ?

Daniel

PS: 1 Inches of water = 248.84 Pa
(so 1 kPa is aprox: 4,12 inches of water)

[Tony]

Great thread Daniel.

There are several important aspects to weigh up when choosing between transducer types.

The first is sensitivity. As 86roccol has pointed out, use the lowest pressure range you can possibly get away with. The electrical output from these sensors is always extremely small, and that means trouble. For instance, if you are only using 5% of the total available pressure measurement range, any temperature drift or zero errors will appear twenty times worse than if you could have used the whole available pressure measurement range.
If measuring only 3Kpa, a 100Kpa range transducer is definitely not a good idea, a 10 Kpa transducer makes a lot more sense.

At the other extreme, these sensors can be permanently damaged by overpressure. If your air blower can reach 120+ inches of completely blocked flow air pressure, one unhappy day, your pressure transducer might see all of that. Make absolutely sure your transducer is rated well above the worst possible overpressure accident that can possibly occur.

The third factor is temperature stability. The manufacturer can do this a lot more accurately than you can. A temperature corrected sensor will cost slightly more, but is absolutely necessary.

Read all the fine print in the specifications. Some transducers can be many times better or worse than others, in various subtle ways.

How many bits a/d resolution ? Realistically ten bits (1024 steps) should be more than sufficient, because the accuracy, noise, and long term stability of the whole system is not going to be better than 0.1% anyway.

A far more fruitful approach to higher pressure resolution is to average a large number of separate readings in software. (statistical averaging) This averages out a lot of the random reading errors due to acoustic noise, electrical noise, and air turbulence. It will provide far more stable high resolution readings than trying to read incredibly small pressure changes directly.

As to differential pressure transducers, the one I really like best so far is the Sensirion SDP2000 discovered by Bruce.

It requires no external amplifier, it can drive an a/d directly with a 4 volt output signal. It is particularly well temperature compensated, and has a full scale pressure range of only 14 inches. It is totally immune to overpressure damage (>400+ inches) because of the highly unusual method of operation. Accuracy is 0.2% and it is available with either a linear or square root output.

These sensors are certainly not cheap at around $90, but they require no circuit board or external amplifier, which offsets some of the high cost.

[MrHijet]

Hi Tony,

Thanks for the long detailed reply. Your link to the sensor was broken, but I know which one you meant ( ).

I am a bit afraid of the costs ( 2x 90 US$) which makes my little project quite expensive. My approach was to get a very cheap display + serial interface. I am looking for a cheaper sensor.

The MPX are temperature compensated, calibrated and even amplified (when using the 5xxx series), so thats really nice.

However you brought me to another idea. To get the noise a little bit down, it would help to stay digital as soon as possible.

I found following manufactor:
(click on "Pressure Sensors and Transmitters" to the the needed overview).

They offer calibrated and digitial pressure sensors. They do the a/d conversion inside the sensor, so the calibration will include the a/d conversion. The communication is possible via SPI and I2C. Thats fine since the Atmel microcontroller are capable of that.

I haven't found prices of they sensors, but I will continue to look around.

Thanks again.

Cheers,

Daniel

[Moriniman]

A couple of thoughts.

The sensor you're looking at (MPX5010) is plus or minus 5% error at full scale. It's also unipolar output, so you would have to physically swap the pressure connections if reversing flow.

Watch out if you use the positive supply rail as your A2D reference voltage, you can lose several bits of usable accuracy due to noise that way.

[MrHijet]

Good hint. I looked around for other sensors with a bidirectional usage, but haven't had luck up to now. However Sensortechnics.com does have such a huge amount of sensors, I hope I will find something (which had a reasonable price) which could fit.

Cheers,

Daniel

[Moriniman]